In recent years, machine learning (ML) has become a key enabling technology for the sciences and industry. Especially through improvements in methodology, the availability of large databases and increased computational power, today's ML algorithms are able to achieve excellent performance (at times even exceeding the human level) on an increasing number of complex tasks. Deep learning models are at the forefront of this development. However, due to their nested non-linear structure, these powerful models have been generally considered "black boxes", not providing any information about what exactly makes them arrive at their predictions. Since in many applications, e.g., in the medical domain, such lack of transparency may be not acceptable, the development of methods for visualizing, explaining and interpreting deep learning models has recently attracted increasing attention. This introductory paper presents recent developments and applications in this field and makes a plea for a wider use of explainable learning algorithms in practice.

We present "AutoJudge", an automated evaluation method for conversational dialogue systems. The method works by first generating dialogues based on self-talk, i.e. dialogue systems talking to itself. Then, it uses human ratings on these dialogues to train an automated judgement model. Our experiments show that AutoJudge correlates well with the human ratings and can be used to automatically evaluate dialogue systems, even in deployed systems. In a second part, we attempt to apply AutoJudge to improve existing systems. This works well for re-ranking a set of candidate utterances. However, our experiments show that AutoJudge cannot be applied as reward for reinforcement learning, although the metric can distinguish good from bad dialogues. We discuss potential reasons, but state here already that this is still an open question for further research.

fMRI semantic category understanding using linguistic encoding models attempts to learn a forward mapping that relates stimuli to the corresponding brain activation. Classical encoding models use linear multivariate methods to predict brain activation (all the voxels) given the stimulus. However, these methods mainly assume multiple regions as one vast uniform region or several independent regions, ignoring connections among them. In this paper, we present a mixture of experts model for predicting brain activity patterns. Given a new stimulus, the model predicts the entire brain activation as a weighted linear combination of activation of multiple experts. We argue that each expert captures activity patterns related to a particular region of interest (ROI) in the human brain. Thus, the utility of the proposed model is twofold. It not only accurately predicts the brain activation for a given stimulus, but it also reveals the level of activation of individual brain regions. Results of our experiments highlight the importance of the proposed model for predicting brain activation. This study also helps in understanding which of the brain regions get activated together, given a certain kind of stimulus. Importantly, we suggest that the mixture of regression experts (MoRE) framework successfully combines the two principles of organization of function in the brain, namely that of specialization and integration.

While deep learning-based classification is generally addressed using standardized approaches, a wide variety of techniques are employed for regression. In computer vision, one particularly popular such technique is that of confidence-based regression, which entails predicting a confidence value for each input-target pair (x,y). While this approach has demonstrated impressive results, it requires important task-dependent design choices, and the predicted confidences often lack a natural probabilistic meaning. We address these issues by proposing Deep Conditional Target Densities (DCTD), a novel and general regression method with a clear probabilistic interpretation. DCTD models the conditional target density p (y x) by using a neural network to directly predict the un-normalized density from (x,y). This model of p (y x) is trained by minimizing the associated negative log-likelihood, approximated using Monte Carlo sampling. We perform comprehensive experiments on four computer vision regression tasks. Our approach outperforms direct regression, as well as other probabilistic and confidence-based methods. Notably, our regression model achieves a 1.9% AP improvement over Faster-RCNN for object detection on the COCO dataset, and sets a new state-of- the-art on visual tracking when applied for bounding box regression. Supervised regression entails learning a model capable of predicting a continuous target value y from an input x, given a set of paired training examples. It is a fundamental machine learning problem with many important applications within computer vision and other domains. While all of these tasks benefit from accurate regression of the target values, high accuracy can even be safety-critical in e.g. automotive and medical applications. Today, such regression problems are commonly tackled using Deep Neural Networks (DNNs), due to their ability to learn powerful feature representations from data. While classification is generally addressed using standardized losses and output representations, a wide variety of different techniques are employed for regression. The most conventional strategy is to train a DNN to directly predict a target y given an input x (Lathuili ere et al., 2019). The training data { (x i,y i)} 2000 i 1 is generated by the ground truth conditional target density p (y x) . DCTD models p (y x) by directly predicting the un-normalized density from the input-target pair (x,y), and is trained by minimizing the associated negative log-likelihood.

Deep learning, through the use of neural networks, has demonstrated remarkable ability to automate many routine tasks when presented with sufficient data for training. The neural network architecture (e.g. number of layers, types of layers, connections between layers, etc.) plays a critical role in determining what, if anything, the neural network is able to learn from the training data. The trend for neural network architectures, especially those trained on ImageNet, has been to grow ever deeper and more complex. The result has been ever increasing accuracy on benchmark datasets with the cost of increased computational demands. In this paper we demonstrate that neural network architectures can be automatically generated, tailored for a specific application, with dual objectives: accuracy of prediction and speed of prediction. Using MENNDL--an HPC-enabled software stack for neural architecture search--we generate a neural network with comparable accuracy to state-of-the-art networks on a cancer pathology dataset that is also $16\times$ faster at inference. The speedup in inference is necessary because of the volume and velocity of cancer pathology data; specifically, the previous state-of-the-art networks are too slow for individual researchers without access to HPC systems to keep pace with the rate of data generation. Our new model enables researchers with modest computational resources to analyze newly generated data faster than it is collected.

Auto-regressive sequence-to-sequence models with attention mechanism have achieved state-of-the-art performance in many tasks such as machine translation and speech synthesis. These models can be difficult to train. The standard approach, teacher forcing, guides a model with reference output history during training. The problem is that the model is unlikely to recover from its mistakes during inference, where the reference output is replaced by generated output. Several approaches deal with this problem, largely by guiding the model with generated output history. To make training stable, these approaches often require a heuristic schedule or an auxiliary classifier. This paper introduces attention forcing, which guides the model with generated output history and reference attention. This approach can train the model to recover from its mistakes, in a stable fashion, without the need for a schedule or a classifier. In addition, it allows the model to generate output sequences aligned with the references, which can be important for cascaded systems like many speech synthesis systems. Experiments on speech synthesis show that attention forcing yields significant performance gain. Experiments on machine translation show that for tasks where various re-orderings of the output are valid, guiding the model with generated output history is challenging, while guiding the model with reference attention is beneficial.

While progress has been made in understanding the robustness of machine learning classifiers to test-time adversaries (evasion attacks), fundamental questions remain unresolved. In this paper, we use optimal transport to characterize the minimum possible loss in an adversarial classification scenario. In this setting, an adversary receives a random labeled example from one of two classes, perturbs the example subject to a neighborhood constraint, and presents the modified example to the classifier. We define an appropriate cost function such that the minimum transportation cost between the distributions of the two classes determines the minimum $0-1$ loss for any classifier. When the classifier comes from a restricted hypothesis class, the optimal transportation cost provides a lower bound. We apply our framework to the case of Gaussian data with norm-bounded adversaries and explicitly show matching bounds for the classification and transport problems as well as the optimality of linear classifiers. We also characterize the sample complexity of learning in this setting, deriving and extending previously known results as a special case. Finally, we use our framework to study the gap between the optimal classification performance possible and that currently achieved by state-of-the-art robustly trained neural networks for datasets of interest, namely, MNIST, Fashion MNIST and CIFAR-10.

In this paper, we propose a novel implicit semantic data augmentation (ISDA) approach to complement traditional augmentation techniques like flipping, translation or rotation. Our work is motivated by the intriguing property that deep networks are surprisingly good at linearizing features, such that certain directions in the deep feature space correspond to meaningful semantic transformations, e.g., adding sunglasses or changing backgrounds. As a consequence, translating training samples along many semantic directions in the feature space can effectively augment the dataset to improve generalization. To implement this idea effectively and efficiently, we first perform an online estimate of the covariance matrix of deep features for each class, which captures the intra-class semantic variations. Then random vectors are drawn from a zero-mean normal distribution with the estimated covariance to augment the training data in that class. Importantly, instead of augmenting the samples explicitly, we can directly minimize an upper bound of the expected cross-entropy (CE) loss on the augmented training set, leading to a highly efficient algorithm. In fact, we show that the proposed ISDA amounts to minimizing a novel robust CE loss, which adds negligible extra computational cost to a normal training procedure. Although being simple, ISDA consistently improves the generalization performance of popular deep models (ResNets and DenseNets) on a variety of datasets, e.g., CIFAR-10, CIFAR-100 and ImageNet. Code for reproducing our results are available at https://github.com/blackfeather-wang/ISDA-for-Deep-Networks.

Neural network models are resource hungry. Low bit quantization such as binary and ternary quantization is a common approach to alleviate this resource requirements. Ternary quantization provides a more flexible model and often beats binary quantization in terms of accuracy, but doubles memory and increases computation cost. Mixed quantization depth models, on another hand, allows a trade-off between accuracy and memory footprint. In such models, quantization depth is often chosen manually (which is a tiring task), or is tuned using a separate optimization routine (which requires training a quantized network multiple times). Here, we propose Smart Ternary Quantization (STQ) in which we modify the quantization depth directly through an adaptive regularization function, so that we train a model only once. This method jumps between binary and ternary quantization while training. We show its application on image classification.

Community detection is a fundamental problem in machine learning. While deep learning has shown great promise in many graphrelated tasks, developing neural models for community detection has received surprisingly little attention. The few existing approaches focus on detecting disjoint communities, even though communities in real graphs are well known to be overlapping. We address this shortcoming and propose a graph neural network (GNN) based model for overlapping community detection. Despite its simplicity, our model outperforms the existing baselines by a large margin in the task of community recovery. We establish through an extensive experimental evaluation that the proposed model is effective, scalable and robust to hyperparameter settings. We also perform an ablation study that confirms that GNN is the key ingredient to the power of the proposed model.